JPS6152979B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a street position detection device on a semiconductor wafer that detects street positions on a semiconductor wafer.

FIG. 1 shows a semiconductor wafer 1 as an example of a detection target. As shown in the figure, a plurality of chips 2 formed on a wafer 1 are divided by streets 3, and in the semiconductor production process, cutting lines are placed on these streets using a scriber, dicer, etc., and the wafer is shaped like a grid. It is cut into individual chips. On each chip 2, as shown in FIG. 2, which is an enlarged view of the portion 4 in FIG. . It is necessary to detect and align the position of semiconductor wafers when testing the wafers using a device called a blower, which inspects the electrical characteristics of each chip. When setting the wafer in a similar device called
This occurs when a wafer is preset in a machine called a printer that prints circuit patterns on the wafer. Now, the apparatus for detecting the streets 3 on the semiconductor wafer 1 will be specifically explained based on FIGS. 3 to 8.

The detection area 7 on the wafer 1 is imaged on a one-dimensional image sensor (line scanner) 26 through an optical system consisting of an objective lens 8 and a cylindrical lens 25, with the light beam being compressed in a certain direction. The position detection circuit 27 calculates the position of the chip 2 from the charge-converted image signal, and based on this detected value, the table 14 is moved via the motor control circuit 12 and the motor drive mechanism 13 to detect the position on the wafer 1. The chip 2 is positioned. The illumination includes a coarse detection illumination 28 that is oblique illumination from a direction perpendicular to the one-dimensional image sensor 26, and a pair of fine detection illumination 29 that is oblique illumination from both left and right directions parallel to the one-dimensional image sensor 26. First, the position detection circuit 27 performs rough detection using the rough detection illumination 28, and then switches to the fine detection illumination 29 to perform fine detection and positioning from the coarse detection value and the image signal. .

FIG. 4 illustrates how a pattern is reflected by oblique illumination using a cross-sectional view of a portion of the wafer 1. Illumination light 30 from diagonally above the left side is regularly reflected by a flat part 31 on the wafer 1 and does not enter the objective lens 8, but it is reflected by the bonding pad part 5, the right edge part 32 of the street, and the step 33 of the circuit pattern. Then, it causes diffuse reflection and enters the objective lens 8. Further, the illumination light 34 coming from diagonally above the right is similarly diffusely reflected by the left edge 35 of the street, the bonding pad 5, and the step 33 of the circuit pattern, and enters the objective lens 8. However, as shown in FIG. 7, the step 36 parallel to the illumination direction causes little diffused reflection, and almost no illumination light enters the objective lens 8. Therefore, by oblique illumination, patterns other than flat areas on the wafer 1 and steps parallel to the illumination direction, wiring patterns, and bonding pads are formed as bright images.

Next, using Fig. 6 and Fig. 7, the coarse detection illumination 2 is
Detection of the position of the chip 2 using the precision detection illumination 8 and the precision detection illumination 29 will be explained. Although it is desired to detect the position of the chip 2 on two axes in the X and Y directions, for the sake of simplicity, only one axis, that is, the position detection in the lateral direction in FIG. 2 will be explained here. In this case, it is sufficient to detect a vertical line in street 3.
FIG. 6a shows a part of the detection area 7 of FIG. As shown in the figure, only the bonding pad 5, the horizontal circuit pattern 37 perpendicular to the illumination, and the street edge 38 are illuminated and detected by the rough detection illumination 28, and the remaining portions are dark and are not detected. When this image is compressed in the vertical direction by the cylindrical lens 25, an image as shown in FIG.
The 0 and street edges 38 are compressed and averaged together with the horizontal circuit pattern 37 and become slightly brighter, but are no longer detected as images. Furthermore, the vertical bonding pad rows 41 are imaged as a bright image 42, and since there is no element of diffuse reflection between the vertical bonding pad rows 41, the image is formed as the darkest image 43. Therefore, Figure 6b
An image signal as shown in FIG. 6c can be obtained from the one-dimensional image sensor 26 on which the one-dimensional pattern shown in FIG. In FIG. 6c, the horizontal direction corresponds to the position of the image, and the vertical direction corresponds to the brightness level, indicating that the top is bright. It is clear that this image signal contains only vertical position information. In this image signal, the lowest level V ₀
shows between vertical bonding pad rows 41 including vertical streets 3. Therefore,
When the signal is cut off at the level of V ₁ which is V ₀ plus an appropriate value DV ₁ , X ₁ and X ₂ corresponding to the inner position of the longitudinal bonding pad row 41 are detected. This X ₁ ,
The middle of X ₂ is set as XC ₁ and is the rough street detection position. That is, in the signal processing by the coarse detection illumination 28, by simply detecting the lowest level V ₀ of the image signal without using complicated pattern recognition,
After that, the coarse center XC ₁ of Street 3 can be found by simple processing. However, since precise position detection is not possible with this illumination, detection is then performed using the precise detection illumination 29.

FIG. 7a shows the state of the pattern by the precision detection illumination 29, and as shown in the figure, only the bonding pad 5, the vertical circuit pattern 44 perpendicular to the illumination, and the edge 45 of the street 3, which is also perpendicular to the illumination, are brightly detected. , other parts are too dark to be detected.
This image is compressed and focused in the vertical direction 39 in the figure by the cylindrical lens 25 and formed on the one-dimensional image sensor 26, as shown in FIG. 7b. In the figure, the horizontal bonding pad array 40 and the vertical circuit pattern 44 are compressed and averaged, but unlike rough detection, a vertical image component 46 remains in the focused one-dimensional image. Further, the vertical bonding pad row 41 and the edges 45 of the vertical streets are brightly focused 47, and the vertical street 3 has the darkest image 48. However, unlike rough detection, there are other dark images 49. The image signal of this one-dimensional focused image is shown in FIG. 7c. The top and bottom indicate the brightness level, with the top being brightest.
Further, the horizontal direction corresponds to the image position. This signal has a complex signal waveform, and it is difficult to detect the signal position corresponding to street 3. but,
If signal processing is performed focusing only on the bright peaks X ₅ and X ₆ sandwiching the coarse center XC ₁ of the street determined by rough detection, it can be easily determined using the following procedure.

First, find an appropriate value for the lowest level V ₂ between X ₅ and X ₆
X ₅ of the video signal at the brightness level V ₃ plus DV ₂ ,
By cutting between _X6 , the signal positions _X3 and _X4 corresponding to edge 45 of the street can be found. This X ₃ ,X ₄
Let the center be the street center XC ₂ . In this way, the vertical street center, i.e.
The lateral position of the chip is detected.

In this method, even if the bonding pad 5 is asymmetrical with respect to the street 3, or even if there are a variety of target products, the street 3 can be easily and reliably determined and the position of the chip 2 can be determined.

FIG. 8 shows an embodiment of the position detection circuit 27.
The main circuit 79 not only controls the position detection circuit 27 but also operates the detection light switching circuit 50 and the motor control circuit 13. First, rough detection light is input. The video signal inputted by the one-dimensional image sensor is sent to a one-dimensional image sensor drive circuit 51 and an amplifier 5.
2. Video memory 5 through A/D converter 53
4 in multilevel gradation. The signals in the video memory 54 go up to the address counter 55 and are read out sequentially, and the lowest value V ₀ is detected by the lowest value detection circuit 56 and stored in the register 57. The threshold value calculation circuit 58 adds the contents of this register 57 and the contents of the threshold value setting register 59 and sets the value in the threshold value register 6.
Store to 0. Next, the threshold value register 60 and the contents of the memory 54 called out by the address counter are compared in the binarization circuit 61, and the high and low values are digitized.
After being converted into a value, it is passed through a differentiation circuit 62. This content is compared with the content coming from the address counter 55 through the delay circuit 63 to determine the edge address and store it in the register 64. Next, the median address of both edges is calculated by the arithmetic circuit 65 and stored in the register 66. Up to this point, the rough detection position of the street has been found.

Next, precise detection is performed in the same loop,
In this case, the range setting address is determined from the rough detection position and the contents of the memory 54 to narrowly limit the scanning range.
X ₅ and X ₆ are detected by a detection circuit and stored in a start address register 67 and an end address register 68, respectively. During rough detection, a start address 69 and an end address 70, which are set in advance, are stored in both registers 67 and 68.

Scanning control is performed as follows.

The difference between these two registers is calculated by the calculation circuit 71 and stored in the scan number register 72. The address counter 55 and the start address counter 74 are started by the set instruction 73, and when the start address counter 74 reaches the contents of the start address register 67, the scanning number counter 7
5, and after counting the contents of the scanning number register 72, a control signal 76 is output and scanning is stopped. As described above, if scanning is performed within a narrow range in the same way as rough detection, fine detection is performed and the contents of the median register 66 are passed to the motor control circuit 12.

In this circuit, a coarse detection threshold difference 77 and a fine detection threshold difference 78 can be stored separately in the threshold setting register 59. Further, logical judgment functions such as error checking are included in the main circuit 79.

As for the illumination method, as for the illumination method, for rough detection, it is sufficient to use dark-field illumination that does not detect vertical edges, and it is not necessary to use illumination from both sides of the pair, but from one direction. good.

Further, in precise detection, it is sufficient to detect edges of the dicing area, and bright field illumination or dark field illumination may be used.

As described above, the conventional street position detection device on a semiconductor wafer optically compresses the reflected light image obtained from a predetermined area using the cylindrical lens 25 and detects it using the photodiode array 10. This method has disadvantages in that it is expensive, the area to be compressed is limited and narrow, and it is difficult to detect the position with high reliability.

It is an object of the present invention to eliminate the above-mentioned drawbacks of the prior art, to make a device for automatically detecting street positions on a semiconductor wafer small and simple, and to be inexpensive, and to totalize reflected light images over a wide area. To provide a street position detection device on a semiconductor wafer which can detect the position of a street on a semiconductor wafer with high reliability.

That is, the present invention includes a first irradiation means that irradiates substantially parallel light onto the semiconductor wafer from obliquely above the longitudinal direction of the street whose position is to be detected; a second irradiation means for irradiating substantially parallel light from the irradiation means; an illumination switching device for switching at least two types of illumination by the first irradiation means and the second irradiation means; and a slit directed in the longitudinal direction of the street. A photoelectric conversion element that detects the amount of diffusely reflected light from the semiconductor wafer and a table on which the semiconductor wafer is mounted are moved in a direction perpendicular to the longitudinal direction of the street to scan the reflected light image with respect to the street. A video signal obtained from the photoelectric conversion element by moving the table by irradiating light with the device, a table movement position detection device for detecting the movement position of the table, and the first irradiation means is the most When the first reference level, which is set slightly higher than the lower level, is reached, the coarse position of the street, which is the midpoint between the two points X ₁ and X ₂ detected by the table movement position detection device, is detected. A rough position detecting means and the irradiation switching device switch from the first irradiation means to the second irradiation means and perform the second irradiation.
With respect to the video signal obtained from the photoelectric conversion element by irradiating light with the irradiation means and moving the table, the set point Precise detection means for detecting the center precise position of the street, which is the midpoint between the two points X ₃ and X ₄ detected by the table movement position detection device when the reference level No. 2 is reached for the first time. To automatically detect the street position on a semiconductor wafer, we use a slit scanning method that uses the movement of the wafer table to avoid the effects of optical system distortion and uneven illumination, resulting in high-quality image detection. By realizing a simple and inexpensive configuration, and by simply changing the illumination method without performing complex electrical processing for image processing, we have realized an easy and inexpensive position detection device. Furthermore, the device uses dark-field illumination and a long slit to expand the total amount of reflected light images, increasing detection reliability.

The present invention will be specifically described below based on embodiments shown in the drawings.

FIG. 9 is a schematic diagram showing an embodiment of a street position detection device on a semiconductor wafer according to the present invention. That is, this device has a pair of left and right detection systems 100, 10.
1, the streets 3 between the chips 2 on the semiconductor wafer 1 are detected, and based on the detection results of 3, the pattern detection positioning is performed to position the wafer 1 in the Y direction and the θ direction at the reference position 81 of the optical system. Next, the wafer 1, which has been aligned in the Y and θ directions, is rotated by 90 degrees and aligned in the X direction in the same manner as above.

Specifically, the detection field 82 on the wafer 1 is enlarged and imaged by the objective lens 8 at the position of a slit plate 84 in which a slit 83 elongated in the X-axis direction is formed.
The amount of light 85 transmitted through the slit 83 opened in the direction is converted into an electrical signal by a photodiode 86,
It is input to the position detection circuit 102. Detection is performed while moving the Y table 14 with the Y drive mechanism 13. The amount of displacement of the Y table 14 is detected by a linear encoder 87 and input to the position detection circuit 102. In the position detection circuit 102, the displacement amount 88 from the linear encoder 87 and the signal 89 from the photodiode 86 are made to correspond to each other, and an image signal 90 corresponding to the pattern on the detection field of view 82 can be obtained. This signal 90 is processed and the position in the Y direction can be detected. The illumination means includes a pair of coarse detection illumination lights 28 that illuminate the semiconductor wafer 1 in a rectangular shape from diagonally above in the X direction perpendicular to the position detection direction, and a pair of coarse detection illumination lights 28 that illuminate the semiconductor wafer 1 in a rectangular shape from diagonally above in the Y direction. It is composed of a precision detection illumination 29 and can be used as follows. That is, first, while moving the wafer 1 using the rough detection illumination 28, the position detection circuit 102 performs rough detection of the street 3. Next, the fine detection illumination 29 is switched to, and the position is determined from the coarse detection value and the image signal 90. The detection circuit 102 performs precise detection of the three street positions. The above detection is performed by a pair of left and right detection systems 100 and 101, and the precise detection values Y ₁ and Y ₂ for the reference position 81 are obtained, and the following (1)
Perform the calculations shown in equations and equations (2).

Y=Y ₁ +Y ₂ /2 ...(1) θ=Y ₁ -Y ₂ /l ...(2) Based on the above formula (1), the motor control circuit 10
3. Move the Y table 14 via the Y drive mechanism 13 and move the θ table 92 via the motor control circuit 103 θ drive mechanism 91 based on equation (2).
Position the semiconductor wafer 1.

Next, detection is performed as shown in FIG. 4 in the conventional manner. That is, FIG. 4 is a partial cross-sectional view showing how a pattern on a semiconductor wafer is reflected by oblique illumination. Illumination light 30 from diagonally above the left
is specularly reflected by the flat part 31 on the wafer 1 and does not enter the objective lens 8, but it causes diffuse reflection by the bonding pad part 5, the right edge part 32 of the street, the step 33 of the circuit pattern, etc., and is reflected by the objective lens. 8. Further, the illumination light 34 coming from diagonally above the right is similarly diffusely reflected by the left edge 35 of the street, the bonding pad 5, and the step 33 of the circuit pattern, and enters the objective lens 8. However, as shown in FIG. 5, the step 36 parallel to the illumination direction causes little diffuse reflection in the direction of the objective lens 8, and the illumination light 30 hardly enters the objective lens 8. Therefore, by oblique illumination, patterns other than flat areas on the wafer 1 and steps parallel to the illumination direction, wiring patterns, and bonding pads are formed as bright images.

Next, using FIG. 10 and FIG. 11, the street 3 is
The position detection will be explained in detail. Figure 10a
is part 4 of FIG. As shown in the figure, only the bonding pad 5 wiring pattern, the Y-direction circuit pattern 51 perpendicular to the illumination, and the street edge 52 are illuminated and detected by the rough detection illumination 28, and the rest of the patterns are dark. Not detected. When the slit 29 moves relative to this image, the position detection circuit 102 uses the displacement amount 95 and the amount of light transmitted through the slit 85 to generate an image signal 9 shown in FIG.
6 is obtained. Y direction bonding pad 40
The edge 38 of the street and the Y direction are averaged by the slit 83 together with the circuit pattern 37 in the Y direction, so although it is somewhat bright, it is no longer detected as an image. Also, the bonding pad row 4 in the X direction
1 is the bright image 97 summed up by the slit 83
Since there is no element of diffuse reflection between the bonding pad rows 41 in the X direction where street 3 exists and there is no amount of light passing through slit 83, it is detected as the darkest image 98. Image signal 9 in FIG. 10b obtained as above
6 does not contain positional information in the X direction because it is averaged by the slit 83, but only contains positional information in the Y direction. In particular, information on long lines, such as Street 3, has a better S/N ratio than information on short lines. This image signal 96
From this, the position detection circuit 102 detects the rough center as follows. In the image signal, the lowest level V ₀ indicates the area between the X-direction bonding pad rows 41 including the X-direction street 3.
When the signal is cut off at the level of V ₁ which is V ₀ plus an appropriate value DV ₁ , X ₁ and X ₂ corresponding to the average position inside the X-direction bonding pad row 41 are detected. this
The middle point between X ₁ and X ₂ is set as XC ₁ and is the coarse detection position of street 3. That is, in signal processing by the coarse detection illumination 28, by simply detecting the lowest level V ₀ of the image signal without using complicated and expensive pattern recognition, the coarse center of the street can be detected with simple processing.
Find XC ₁ . However, since detailed position detection cannot be performed with this coarse detection illumination, next we will use the fine detection illumination 2.
9 is performed.

FIG. 11a shows the state of the pattern produced by the precision detection illumination 29, and as shown in the figure, only the bonding pad 5 circuit pattern, the X-direction circuit pattern 44 perpendicular to the illumination, and the street edge 45 also in the X direction are shown. is detected brightly, and other parts are dark and therefore not detected. Slits 83 for this statue
When there is a relative movement, the position detection circuit 102 obtains an image signal 99 shown in FIG. 9B from the displacement amount 95 and the amount of light transmitted through the slit 85. In the figure, the Y-direction bonding pad array 40 and the X-direction circuit pattern 44 are averaged by the slit 83, but unlike rough detection, the X-direction circuit pattern 44 has a long line length, so it is detected as an image. In addition, the X-direction bonding pad row 41 and the X-direction street edge 45
is brightly focused by the slit 83, and the street in the X direction becomes a dark image 105. However, unlike rough detection, other dark images 106 are also produced. This image signal 9
9 is FIG. 11b. The horizontal direction of the figure corresponds to the pattern position, and the vertical direction corresponds to the brightness, with the right direction indicating brightness. This signal 99 has a complex signal waveform, and it is difficult to easily detect the signal position corresponding to street 3. but,
If signal processing is performed focusing only on the bright peaks X ₅ and X ₆ sandwiching the coarse center XC ₁ of the street determined by rough detection, the accurate center of the street can be easily determined using the following procedure.

First, find an appropriate value for the lowest level V ₂ between X ₅ and X ₆
X ₅ of the image signal at the brightness level V ₃ plus DV ₂ ,
By cutting between _X6 , the signal positions _X3 and _X4 corresponding to edge 45 of the street can be found. This X ₃ ,X ₄
Let the center be the street center XC ₂ . In this way, the street center in the X direction was determined.

In this method, even if the bonding pad 5 is asymmetrical with respect to the street 3, or even if there are many types of products, the street can be easily and reliably determined. Furthermore, since the length of the slit is longer than the length of the chip on the semiconductor wafer 1, the edge of the street can be detected with good S/N ratio.

As explained in detail above, regarding the illumination method, for coarse detection, it is sufficient to use dark field illumination that does not detect vertical edges, and illumination from one direction is sufficient instead of illumination from both sides of the pair. .

Further, in precise detection, it is sufficient to detect the edges of the street, and bright field illumination or dark field illumination may be used.

The input is not limited to a photodiode, and any photoelectric conversion element such as a photomultiplier may be used as long as it has an output proportional to the amount of light.

By performing rough positioning on the basis of rough detection values before fine detection, the table scanning distance is shorter than in the previous embodiment, and therefore the detection time can be shortened. Furthermore, since the θ direction is corrected, if the optical system reference position matches the X direction, the slit direction and pattern direction match during precise detection, and an image signal with high resolution can be obtained.

As explained above, according to the present invention, streets on a semiconductor wafer can be detected with high accuracy by improving detection sensitivity, and the street position detection device can be made inexpensive by simplifying the detection system. be effective.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing a semiconductor wafer as an example of a detection target, Fig. 2 is a partially enlarged view of Fig. 1 to explain the detailed pattern, and Fig. 3 is an example of a conventional street position detection device on a semiconductor wafer. Fig. 4 is a diagram showing how oblique illumination light is reflected by a pattern, Fig. 5 is a diagram showing how it is reflected by a pattern parallel to oblique illumination light, and Fig. 6a is a rough detection diagram. Diagram showing how the pattern looks depending on lighting,
b is a diagram showing a compressed condensed image of that image by a cylindrical lens, c is an image signal waveform diagram of the compressed condensed image, FIG. 7a is a diagram showing how the pattern looks with rough detection illumination, and b is a diagram A diagram showing a compressed condensed image of that image by a cylindrical lens, c is an image signal waveform diagram of the compressed condensed image, FIG. 8 is a diagram concretely showing the position detection circuit shown in FIG. FIG. 9 is a schematic configuration diagram showing an embodiment of the street position detection device on a semiconductor wafer according to the present invention, and FIG. 10a shows how the pattern is viewed by the coarse detection illumination of the device shown in FIG. 9, and the positional relationship of the slits. Figure 11b is a diagram showing the image signal waveform obtained by scanning the image with the slit, and Figure 11a is a diagram showing how the pattern looks with the precise detection illumination of the apparatus shown in Figure 9 and the slit. A diagram showing the positional relationship, and b is a diagram showing the image signal waveform obtained by slit scanning of the image. Explanation of symbols: 1... Wafer, 2... Chip, 3... Street, 8... Objective lens, 13... Y drive mechanism,
14...Y table, 28...Rough detection illumination, 29...
Illumination for precise detection, 83...Slit, 84...Slit plate, 86...Photodiode.

Claims (1)

[Claims] 1. A first irradiation means that irradiates substantially parallel light onto the semiconductor wafer from diagonally above the longitudinal direction of the street whose position is to be detected; a second irradiation means for irradiating at least two irradiation means, and at least two
a lighting switching device for switching between types of lighting; a photoelectric conversion element for detecting the amount of diffusely reflected light from the semiconductor wafer through a slit extending in the longitudinal direction of the street; and a stage for mounting the semiconductor wafer in the longitudinal direction of the street. a device that scans the reflected light image with respect to the street by moving the platform in a direction perpendicular to the street; a platform movement position detection device that detects the movement position of the platform; and a device that irradiates light with the first irradiation means. When the video signal obtained from the photoelectric conversion element by moving the stage reaches a first reference level set slightly higher than the lowest level, the stage is detected by the stage movement position detection device.
coarse position detection means for detecting a rough position of the street that is an intermediate point between points X ₁ and X ₂ ; and switching from the first irradiation means to the second irradiation means by the irradiation switching device. A video signal obtained from the photoelectric conversion element by irradiating light with the second irradiation means and moving the table is set across the rough position of the street detected by the rough position detection means. Precise detection means for detecting a precise center position of the street, which is an intermediate point between the two points X ₃ and X ₄ detected by the table movement position detection device when the second reference level is first reached; A street position detection device on a semiconductor wafer, characterized in that it is equipped with a.